DE102020007273A1 - Method and device for the automatic detection and classification of acoustic signals - Google Patents

Abstract

Method relates to automatic detection and classification of acoustic signals with at least one acoustic signal receiver arranged in the monitoring area and a module for classifying the acoustic signal and outputting classification information. The device for carrying out the method comprises at least one acoustic sensor system (5), a raw acoustic database (10) of the acoustic sensor system (5), interfaces (S1 to S6) with an A / D converter (2), a hardware module motion detector (3), a hardware module GPS (4), a Hardware module video camera (5), a hardware module wind sensor (6) and a hardware module rain sensor (7) are connected, an acoustic model database (11) in which acoustic models are stored and a classification module (8) with an integrated acoustic background signal recording module, which is connected to the acoustic model database (11) and to an evaluation application (13) and in the event that a signal is detected, the classification result is sent to the evaluation application (13), so that the acoustic sensor system (S) can be used both as a recording system for deviations in the acoustic background for a new training phase and as a classification module for known signal patterns during is used for a classification phase and generates the recognition models required for this from the signals previously recorded by the acoustic sensor system (S) and acoustic background data that are different individually in each sensor system.

Description

The invention relates to a method and a device for the automatic detection and classification of acoustic signals in a surveillance area to prevent vandalism and crime in objects through to the monitoring and early detection of faults in machines, installations and systems.

Differently designed methods and devices for the automatic detection and classification of acoustic signals in a surveillance area are known in various technical fields. Monitoring systems for monitoring an open space or an area in a building or building complex, the monitoring system comprising a plurality of microphones which can be and/or are arranged in the monitored area and which are designed to record audio signals of an object as an audio signal source in the monitored area known for a long time.

From the DE 10 2012 211 154 A1 a monitoring system is known which comprises an analysis module for classifying the audio signal and for outputting classification information. The classification information is in particular a description of the content of the audio signal and describes the type of noise. In particular, the classification information designates the type of origin of the audio signal. Furthermore, an analysis module is provided, which compares an audio signal that is present with a pattern and/or a stored audio signal. In particular, the analysis module is designed to classify an audio signal as an unintended noise, namely as a burglary noise and/or as a damage noise and/or as an interference noise of an action of the object as a type of noise.

The surveillance system targets the unintended noises that people either cannot prevent (sounds of footsteps or sounds of an action) or accidentally do not sufficiently prevent. The surveillance system includes a localization module for locating the position of the object by acoustically locating the position of the audio signal source of the classified audio signal in the surveillance area via the microphones and for outputting position information. The localization module is designed to localize the person of the audio signal source of the classified audio signal by acoustic cross-location of the audio signal source, in particular by measuring the propagation time of the audio signal from at least two and in particular at least three microphones.

The classification information and the position information form part of an object data record, wherein the object data record can be output and/or further processed as a monitoring result. For this purpose, the monitoring system includes an evaluation module for evaluating the object data set and for generating an alarm signal, it also being provided that the evaluation module uses at least two object data sets with different classification information for the evaluation. In one development, the evaluation module has a tracking unit, which is designed to create a metadata record for the object from a number of object data records. In this metadata record, for example, two, three, four or more possible object data records of the object are recorded and the evaluation module includes a filter unit for filtering the metadata record, with object data records that contradict a movement model of the object being filtered out. Finally, the monitoring system includes a monitoring center which has an audio storage device or is connected to it, an audio recording of the audio signal, in particular the original audio signal of the classified audio signal, being stored on the audio storage device.

Furthermore, from the WO 2009/104968 A1 a burglar alarm system is known in which not the audible signal but the inaudible signals are evaluated in order to determine the cause of the noise. For this purpose, filters are used to separate the low-frequency signal (frequency range of around 1-5 Hz) and high-frequency signal (frequency range of around 5-20 Hz) and a comparator compares the signals in the two channels with one another.

From the DE 10 2014 012 184 B4 a device and a method for the automatic detection and classification of acoustic signals in a surveillance area are known. In order to enable all signal-influencing factors to be included in the respective recognition model, an acoustic sensor system with a microcomputer is provided for pre-processing the acoustic signals, which is connected to a signal receiver whose output signals are parallel to a module arranged in the acoustic sensor system for their recording and a Module for classification are supplied. A recording database of the acoustic sensor system is connected to the recording module, in which the signal is stored in the format of an audio file. For data exchange and control with an interface having an acoustic sensor system is a module for modeling in connection, which has a Interface imports the recordings, generates corresponding models from them and which is connected via the model library of a corresponding sensor system, in which the model selected by a user in a training phase is stored. The classification module is connected to the model library and the classification module is connected to an evaluation application via a further interface, and if a signal is detected, sends the classification result to the evaluation application.

the DE 10 2017 012 007 A1 describes a device and a method for universal acoustic testing of objects, the device for acoustic testing of objects having a striking tool, a free-standing microphone and/or a structure-borne microphone for detecting the sound and with an analysis device connected thereto for analyzing the sound up to the ultrasonic range included in. The device is designed as a device which has an excitation device mounted outside of a device housing and an acoustic sensor arranged adjacent thereto, and the output of the acoustic sensor can be connected to either an analysis module or a training module inside the device housing by means of a manually operated operating mode switch, wherein the output of the training module is connected to a model library which is connected to the analysis module.

As the above prior art shows, differently designed methods and devices for the automatic detection and classification of acoustic signals in a surveillance area are known in various technical fields.

As a rule, the signals are sent to a central location for recording and for comparison with a reference pattern previously stored in a library, which leads to considerable effort in recording and classification (fixed model-based reference classification or learned reference classification from the previous classifications). leads. In practice, therefore, there is a lack of cost-effective methods and devices with regard to the incorporation of all signal-influencing factors into the respective recognition model, in particular taking into account the current situation and simple configuration or adaptation in a training phase by the user.

The factors influencing the relevant signals are not taken into account or are incorrectly taken into account in the known prior art, and the user also makes subjective interventions in this important part of the acoustic model generation in a training phase. This is more or less goal-oriented from user to user and actually unnecessary.

The invention is based on the object, based on known methods and devices for the automatic detection and classification of acoustic signals in a surveillance area, to design them in such a way that when configured by the user, all signal-influencing factors are included in the respective detection model and all device-related signals are influencing Factors such as components and assemblies, which are subject to manufacturing tolerances, site-related factors, such as the site-related individually different acoustic background, together with an acoustic signal preserve that is the same for all sensors for each noise to be detected, are included in the respective detection model.

The object is achieved by a method with the features according to patent claim 1 and by a device with the features according to patent claim 7. Further configurations are shown in the dependent claims.

According to the invention, at least one acoustic sensor system, which has a microcomputer and is connected to an acoustic signal receiver, is provided for preprocessing the acoustic signals. Compares the model library and, in the event of a match, stores it in the acoustic signal detection database (depending on the type of use of the sensor for further processing and alarming locally in the sensor system itself or centrally outside of the sensor system and, on the other hand, carries out a permanent analysis of the stationary acoustic background and the detected acoustic deviations in the acoustic signal raw data database, whereupon after a configurable number of such deviations in the acoustic background the module updates the model automatically and without any action being required rch the user updates an acoustic signal model database, after which the module for classification works with an acoustic background based on the currently prevailing acoustic conditions and henceforth carries out its classifications with this updated acoustic model, that the classification module with the Model library is connected and that the classification module is connected to an evaluation application via a further interface, and in the event that a signal is detected, the classifi The evaluation result is sent to the evaluation application, so that the acoustic sensor system can be used as a recording system only for any acoustic signals of the stationary, different acoustic background of the sensor in preparation for a training phase for a new acoustic model or one that is to be updated, as well as known signal patterns as a classification module serves during a classification phase and generates the recognition models required for this from previously recorded signals that are the same for all sensors and the respective different acoustic background signals of the respective sensor, and these models are then only used by this sensor system for classification.

The output signals of the acoustic signal receiver of the acoustic sensor system are fed to a classification module with an integrated acoustic background recording module, which generates or inserts a new acoustic model in a training phase from these acoustic background signals and from the acoustic signals that are the same for all sensors existing acoustic model is updated and in the classification phase these output signals are classified and compared with already known signal patterns.

A corresponding recognition model is automatically generated by the module for modeling model update and made available to the classification module. Each acoustic sensor system is provided with the same recordings of the acoustic signals to be recognized and enriched with its own recordings of the different acoustic background for each sensor. The resulting models that are generated are then only used for this sensor for classification.

A model update cycle runs automatically within the model update module on the basis of data from the classification module in a loop and the acoustic models created by this cycle become the previously known ones by adding the stationary acoustic background deviations formed acoustic background of the respective sensor per sensor system and used by the respective sensor system.

An arranged module for modeling model update is connected to the acoustic sensor system. In the sensor system, the two modules for classification and model creation/model update can be activated and deactivated from the management application. This makes it possible to use the two operating modes training and classification separately or to operate both in parallel.

However, the acoustic signal to be recognized is recorded beforehand, without any acoustic background signals and at different application-related distances between the sensor and the acoustic signal source, but with the same acoustic signal receiver used later in the sensor and stored in the acoustic signal raw data database. As a result, these acoustic signals to be recognized are the same for each sensor of the device described here and do not have to be newly generated and recorded there in each case. In addition, within the scope of this invention, there is the possibility of further enriching the acoustic signals to be recognized at a central point for each sensor, if necessary.

The device for performing the automatic detection and classification of acoustic signals according to the method according to the invention comprises at least one acoustic sensor system, a recording database of the acoustic sensor system, interfaces connected to an A/D converter, a hardware module motion detector, a hardware module -GPS, a hardware module video camera, a hardware module wind sensor and a hardware module rain sensor, an acoustic model database in which previously acoustic models are stored and a classification module with an integrated background signal -Recording module, which is connected to the acoustics model database and to an evaluation application and, in the event of a signal being detected, sends the classification result to the evaluation application, so that the acoustic sensor system can also be used as a recording system for deviations in the acoustics -Background for a training phase al s also serves as a classification module for known signal patterns during a classification phase and generates the necessary recognition models from the signals previously recorded by the acoustic sensor system and individually different acoustic background signals in each sensor system and these models are then only used by this acoustic sensor system for classification will.

An acoustic sensor system with a microcomputer is provided for pre-processing the acoustic signals compares and, in the event of a match, saves the output signals in the corresponding acoustic signal detection databases of the sensors for further processing and alarming and, on the other hand, a permanent analysis of the stationary acoustic background and stores the detected acoustic deviations in the acoustic signal raw data database, whereupon after a configurable number of such deviations in the acoustic background the module updates the acoustic signal model database automatically and without any action required by the user, after which the Classification module works with the currently prevailing acoustic conditions of the acoustic background and carries out its classifications with this updated acoustic model from now on.

An acoustic sensor system with a microcomputer is provided for pre-processing the acoustic signals is stored in the format of an audio file, and is connected to a module for modeling, which imports recordings from the recording database, generates corresponding models from them and is connected to an acoustic model database of the corresponding sensor system, in which the acoustic model is stored, and that the classification module is connected to the acoustic model database and via a further interface to an evaluation application, and if a signal is detected, the classification result to the evaluation application n transmits, so that the acoustic sensor system serves both as a recording system for detected acoustic background deviations for a training phase and as a classification module for known signal patterns during a classification phase and generates the necessary recognition models from previously recorded signals and acoustic background signals and these models are then only used by be used for classification with this sensor system.

The model update module for model creation and updating is connected to the acoustic sensor systems and activates or deactivates both the classification module completely after the model update has taken place in the form of its restart and during an acoustic model update that is being processed with regard to deactivation /Activation of further acoustic recordings of deviations in the acoustic background of the respective sensor during and after the acoustic model update.

The sensor system has an amplifier connected to the acoustic signal receiver, whose amplified audio output signal is fed to an analog-to-digital converter and that a microcomputer is connected both to the A/D converter and to at least one interface of the interfaces for data exchange and control of the sensor system is connected via a network.

The sensor system has an amplifier connected to the acoustic signal receiver, an analog-to-digital converter and a microcomputer, which can be connected both to the amplifier and analog-to-digital converter via a first interface and to at least one other interface for controlling other connected sensor systems ( Motion detector, GPS, video camera, wind and rain sensor depending on the operating mode) are connected by the sensor system via the network.

For example, capacitive converters (condenser microphones), electrodynamic converters (dynamic microphones, pick-ups) or piezoelectric converters can be used as acoustic signal receivers.

The device according to the invention has the advantage that all relevant components for signal processing and classification are located in an intelligent device, the acoustic sensor system. In the method according to the invention, the sensor system serves as a recording system for acoustic backgrounds that are different depending on the location of the sensor and for the recognition acoustic signal recordings previously stored in the sensor system and not to be recorded again for a training phase, as well as a classification module for already modeled and therefore known signal patterns during the classification phase .

During the classification phase, the model parameters of the underlying recognition model are adapted by the current acoustic input signals. This, but above all the permanent and automatic acoustic model update by the model update module, means that the recognition models are constantly and above all automatically optimized in the course of operation without the user having to do anything. The acoustic sensor thus learns completely independently and thereby enables or improves the quality of the classification under a wide variety of acoustic conditions.

A decisive advantage of the invention compared to the current state of the art is that each acoustic sensor system detects the same tion-acoustic signal data and his own recordings regarding the different acoustic backgrounds of each sensor system for modeling and this model is then only used by this device for classification. This ensures that all factors influencing the signals are included in the respective recognition model and the configuration in the training phase is made possible for the user in a surprisingly simple manner.

Above all, the permanent and automatic acoustic model update by the model update module means that the recognition models are constantly and above all automatically optimized in the course of operation without the user having to do anything. The acoustic sensor thus learns completely independently and thereby enables or improves the quality of the classification under a wide variety of acoustic conditions.

Further advantages and details can be found in the following description of a preferred embodiment of the invention with reference to the drawing 1 remove.

the 1 shows the network structure and an overview of all components of the device according to the invention, the general structure of a network of acoustic sensors according to and an embodiment for carrying out the method according to the invention.

the 1 shows an overview of all components of the device according to the invention. The acoustic signal receiver 1, in particular a microphone, receives an acoustic signal and forwards this to an acoustic sensor system S via a connected amplifier and A/D converter 2. In the sensor system S, the signal is fed to a classification module 8 for classification. A recording module integrated into the classification module 8 has the task of carrying out a permanent analysis of the stationary acoustic background and storing the acoustic deviations found in the raw acoustic database 10 .

An A/D converter 2, a hardware module motion detector 3, a hardware module GPS 4, a hardware module video camera 5, a hardware module wind sensor 6 and a hardware module rain sensor 7 connected to sensor microcomputer S-MC. Furthermore, the sensor microcomputer S-MC is connected to the management application 14 and the hardware module power supply 15 .

A central unit Z is also connected to the sensor system S via a computer network 16 . The central unit Z contains a central microcomputer Z-MC, which includes a management application 14, an evaluation application 13 and an acoustics recognition database 17, with a hardware module power supply 15, a hardware Module wind sensor 6 and a hardware module rain sensor 7 is connected.

The classification module 8 is connected to a model library 11 for its task of classifying the input signal. The acoustic model database 11 includes the acoustic models of all known signals. If there is no acoustic model database 11, the classification module 8 cannot work. To generate an acoustics model database 11, the audio recordings of the raw data target noises previously generated for the sensor system S from the raw acoustics database 10 are used.

The module for modeling/model update 9 imports the audio recordings for the target sounds to be recognized from the acoustic raw database 10, generates corresponding acoustic models from them and the corresponding background recordings and places them in the acoustic model database 11 of the corresponding sensor system S. Each sensor system S has its own model update module 9. If the classification module 8 has detected an acoustic signal, it sends the classification result to a management application 14. This application 14 is responsible for further processing of this information and can take appropriate actions , such as B. alarm or control tasks.

In the sensor system S, the two classification modules 8 and model update modules 9 can be activated and deactivated from the management application 14 . This makes it possible to use the two operating modes training and classification separately or to operate both in parallel. Due to the parallel operation of both modules, it is possible in the classification phase to also record deviating and previously unknown background signals to the system, which can later be used to improve existing acoustic models in the acoustic model database 11 with regard to the expand detection.

The lower part of 1 shows the general structure of the centralization and networking of acoustic sensor systems S. In detail, it has the sensor system S, which is connected to the acoustic signal receiver 1, des sen output signals are fed to a classification module 8, which on the one hand classifies the output signals and compares them with existing acoustic models from the acoustic model database 11 and, in the event of a match, into the acoustic recognition database 12 and/or 17, depending on the operating mode, on the other Processing and alarming stores and on the other hand carries out a permanent analysis of the stationary acoustic background and stores the detected acoustic deviations in the acoustic raw database 10.

If the sensor system S does not work in single-sensor mode, all detection results are stored in the acoustic detection database 17 via the computer network 16. The basis for networking a sensor domain is a classic local computer network LAN 16 or the wireless variant WLAN, both modules are integrated components of the microcomputer S-MC.

Another interface built into the microcomputer is S4. It is used to output video signals to a video camera or to an HDMI device such as a B. for connecting a monitor for configuration directly on the sensor system S. The external components evaluation application 13 and management application 14 use standard network protocols for data exchange and for controlling the acoustic sensor systems S using the microcomputer S-MC, which meets the requirements for networking via LAN or WLAN provides. Entire sensor domains can be realized with this type of networking. In the domains, the individual sensor systems S work completely independently, but can be configured and controlled using classic network functions.

In the method according to the invention, the recognition models are generated from acoustic raw data signals that are previously recorded independently of the sensor and are the same for all sensor systems S. From these, together with the individually location-dependent, different background recordings, a corresponding recognition model is then automatically generated in each sensor system S and made available to the classification module 8 .

During a training phase for model creation or model updating, the analysis and further recording of deviating background acoustic data in the classification module 8 is deactivated until the new/updated acoustic model is provided by the model update module 9 . This prevents further background recordings, because these are most likely already included in the current model update.

In detail, a management application 14 is available for the administration and management of the sensor systems S. With this application, the user can view or change the configuration of each sensor system S, as well as the programs and services running on the sensor system S. It is also possible to view the status of the existing acoustic models from here. The configuration of each sensor system S is stored in a separate area and table in the acoustic model database 11 .

The use of the management application 14 takes place via a previously created bot account of a messenger provider, e.g. the provider Telegram or Whatsapp. Both provider bot api's are supported by the sensor system S.

1. a) die im Sensor aktiven Akustik-Modelle im Klartext unter „Isnoise0...n“ sowie deren direkt zugeordnete Parameter „matchth“ - zur Eintragung in die Akustik-Erkennungsdatenbank (12 und/oder 17) je nach Betriebsart notwendige Anzahl von Treffern pro Erkennungssegment
2. b) den für das jeweilige akustische Erkennungssignal oder alle akustische Erkennungssignale einheitliche minimale „lo_freq“ und maximale „hi_freq“ Frequenz, um das Frequenzspektrum für die Klassifizierung festzulegen
3. c) die Sample Frequenz in „samplingrate“ in Hz, das Abtastintervall in „samplingtime“ in s.
4. d) unter „Ismainpath“ der Projektpfad auf dem Sensorsystem S.

Several acoustic sensor systems S can be operated in parallel in an acoustic classification system. Each individual sensor system S works with a recognition model specially calculated for it. The configuration of each sensor system S can be viewed or changed within the management application 14 by entering the “config” command, such as in the “soundlv” area:

1. a) The acoustic models active in the sensor in plain text under "Isnoise0...n" and their directly assigned parameter "matchth" - for entry in the acoustic recognition database (12 and/or 17) depending on the operating mode, the number of hits per recognition segment
2. b) the uniform minimum "lo_freq" and maximum "hi_freq" frequency for the respective acoustic detection signal or all acoustic detection signals in order to define the frequency spectrum for the classification
3. c) the sample frequency in "sampling rate" in Hz, the sampling interval in "sampling time" in s.
4. d) under "Ismainpath" the project path on the sensor system S.

A project is understood to mean a set consisting of 1 to n intelligent sensor systems S. The sensor systems S are managed within projects. All sensor systems S are created and administered in the acoustic detection database 12 and/or 17 in MySQL database tables, e.g. the “Sensor” table, depending on the operating mode of the sensor.

A customer ID and location ID assigned to each sensor system S there assigns each Sen assign a customer and a location. With the sensor operating mode "single-sensor operation" this is only one at a time, with networks the number can theoretically be n.

Each sensor system S must be entered in the acoustic detection database 12 and/or 17, depending on the operating mode of the sensor -> single-sensor operation or -> multi-sensor operation), otherwise no database action takes place, e.g. E.g. alarm in the event of a positive detection of an acoustic signal. The entries are made using the MySQL database interface.

The Hidden Markov Model (HMMI) on which the known invention is based and used for generating acoustic signal models is deliberately not used by this invention, since the Hidden Markov Model (HMMI) generates acoustic signal models in addition to all recognizing acoustic signal models of the target noises to be recognized, the acoustic background of the sensor system S must also be stored as a recognition acoustic model and the system then makes a selection from all these acoustic models when classifying/recognizing an acoustic signal . This presupposes that, in particular in the acoustic background model, all variants of possible acoustic backgrounds must be observed in order to be able to make an almost error-free selection in the classification / recognition.

However, since it is impossible to pack all possible acoustic background variants into one acoustic model and the misidentifications to be expected in practice must therefore occur, the entire classification in the present invention was implemented using a proprietary classification solution, which exclusively uses the acoustic signals to be classified according to the acoustic modeling for the classification and the individually location-dependent acoustic background of each sensor only for the change analysis of the background in the acoustic model database 11 .

An individual model is created for each sensor system S, based on which the sensor system S then works. This requires one or more corresponding training phase(s) in order to also individually configure the sensor system S and adapt it to the place of use. This is done fully automatically by the classification module 8 and the model update module 9 on the basis of the configuration data stored by the user.

1. a) „maxbrecfc“ legt fest, wie viele Akustik-Hintergrundsegmente in s aufgezeichnet worden sein müssen, um eine Akustik-Modellaktualisierung zu starten;
2. b) „noisemaxten“ in s legt fest, wie viel Akustik-Hintergrund-Aufzeichnungszeit pro Hintergrund-Akustik-Modell gespeichert werden sollen, bis die ältesten Aufzeichnungen wieder überschrieben werden;
3. c) „minmatchthbg“ legt fest, ab welcher minimal erkannten Anzahl von Treffern pro Abtastintervall der bis dato bekannte Akustik-Hintergrund erkannt werden muss, um als bekannt zu gelten;

Values stored in the configuration in the "mupdate" area, such as:

1. a) "maxbrecfc" specifies how many acoustic background segments must have been recorded in s to start an acoustic model update;
2. b) "noisemaxten" in s specifies how much acoustic background recording time should be saved per background acoustic model until the oldest recordings are overwritten again;
3. c) “minmatchthbg” defines the minimum number of hits per sampling interval from which the hitherto known acoustic background must be recognized in order to be considered known;

As a rule, this value is chosen to be very small, e.g. 1, whereby the type of modeling of the acoustic signals to be recognized already ensures that no acoustic signal to be recognized is classified as an unknown background segment. This threshold represents an additional safeguard to prevent this.

Correspondingly higher threshold values are also stored for the acoustic signals to be recognized. In addition, an internal logic in the classification module 8 ensures that when one of the target acoustic signals is detected, the background is not analyzed for this time. With all these measures and automatisms, a permanent and automatic acoustic model update in each sensor system S is made possible.

Since each sensor system S contains an integrated microcomputer S-MC, the sensor systems S are coupled to one another and to the management 14 and evaluation applications 13 via a computer network (LAN) 16 or a wireless network (WLAN). This type of coupling is technically very mature and can be implemented economically.

The acoustic signal receiver 1, z. B. microphone or sound transducer are connected directly to the sensor system S or are mechanically (e.g. with a gooseneck) and connected by audio cables in the immediate vicinity. Furthermore, it is possible to arrange the acoustic signal receiver 1 in the housing of the intelligent sensor system S in such a way that it can be adjusted vertically and laterally towards the monitored area. It is equally possible to provide the acoustic signal receiver 1 for outdoor use with a wind shield in order to reduce or completely avoid clipping or noise caused by wind.

The signal processing in the acoustic sensor system S is largely independent of the acoustic signal receiver 1 used. However, the generated acoustic models differ depending on the acoustic signal receiver 1 used and the sen mechanical installation in a housing (e.g. windbreak) from each other. For this reason, the data for the sensor system S in the raw acoustic database 10 must always have been generated with the same acoustic signal receiver 1 and its mechanical structure.

However, this fact does not represent a major disadvantage, since the acoustic signal raw data is recorded loss-free and unchanged in environments with little or no background noise and can then be recorded by all different types of signal receivers. In this way, the generation of raw acoustic signal data for a new type of signal receiver can be repeated at any time and the recording can then be stored in the raw acoustic database 10 of the corresponding sensor system S.

• • Echos, hervorgerufen durch Laufzeiten, Geometrie und Beschaffenheit des Raumes, in welchem sich der Sensor befindet;
• • Typ und Toleranz der akustischen Signalempfänger (z. B. Mikrofone) und Signalverarbeitung.
• • Ortsabhängige Pegel der Eingangssignale;
• • stationäre Hintergrundgeräusche.

Eventuelle nachteilige Effekte durch eine jeweilige akustische Erkennungssignalerzeugung am Einsatzort des Sensorsystems werden dadurch vermieden. Somit ist diese Erfindung sehr variabel und ortsungebunden einsetzbar!A decisive advantage of the invention compared to the current state of the art is that each acoustic sensor system S uses the same previously recorded background-noise-free acoustic data and its own background acoustic data for acoustic modeling and this model is then only used by this device for classification. As a result, all signal-influencing factors flow into the respective recognition model. Factors influencing the signal are:

• • Echoes caused by propagation times, geometry and nature of the room in which the sensor is located;
• • Type and tolerance of acoustic signal receivers (eg microphones) and signal processing.
• • Location-dependent levels of the input signals;
• • stationary background noise.

Any disadvantageous effects caused by a respective acoustic detection signal generation at the place of use of the sensor system are thereby avoided. Thus, this invention is very variable and can be used anywhere!

Due to a permanent acoustic model update, automatically triggered by a changed acoustic background, the invention described here has real practicability, both indoors and outdoors for objects and systems to be monitored.

Reference List

1 -

Acoustic signal receiver

2 -

A/D converter with integrated amplifier

3 -

Hardware module motion detector

4 -

Hardware module GPS

5 -

Hardware module video camera

6 -

Hardware module wind sensor

7 -

Hardware module rain sensor

8th -

Classification Module

9 -

Model update module

10 -

Acoustics raw database

11 -

Acoustics Model Database

12 -

Acoustic detection database

13 -

evaluation application

14 -

management application

15 -

Hardware module power supply

16 -

computer network

17 -

Acoustic detection database

S -

sensor system

S-MC -

sensor microcomputer

Z -

central unit

Z-MC -

Central microcomputer

S1, S2, S3, S4, S5, S6 -

interfaces

QUOTES INCLUDED IN DESCRIPTION

This list of the documents cited by the applicant was generated automatically and is included solely for the better information of the reader. The list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions.

Patent Literature Cited

• DE 102012211154 A1 [0003]
• WO 2009/104968 A1 [0006]
• DE 102014012184 B4 [0007]
• DE 102017012007 A1 [0008]

Claims (9)

Method for the automatic detection and classification of acoustic signals with at least one acoustic signal receiver arranged in the monitoring area and a module for classifying the acoustic signal and outputting classification information, characterized in that at least one acoustic sensor system (S) having a microcomputer is provided for preprocessing the acoustic signals which is connected to an acoustic signal receiver (1), the output signals of which are fed to a classification module (8), which on the one hand classifies the output signals and compares them with existing acoustic models from the acoustic model database (11) and in the event of a match, stores it in the acoustic signal recognition database for further processing and alarming and, on the other hand, carries out a permanent analysis of the stationary acoustic background and the detected acoustic deviations in the acoustic Raw database (10) saves, whereupon after a configurable number of such deviations in the acoustic background, the model update module (9) automatically and without any action required by the user in an acoustic model database (11) updated, after which the module for Classification (8) works with one on the currently prevailing acoustic conditions of the acoustic background and henceforth carries out its classifications with this updated acoustic model, with an acoustic model database (11) of the corresponding sensor system (S) being used by a user model selected in a training phase is stored, that the classification module (8) is connected to the acoustic model database (11) and that the classification module (8) is connected to an evaluation application (13), and in the event that a signal is detected, the classification result is sent to the evaluation application (13), so that the acoustic Sen sensor system (S) serves both as a recording system for any acoustic signals during a training phase and as a classification module (8) of known signal patterns during a classification phase and generates the necessary recognition models from the signals previously recorded by the sensor system (S) and these models then only used by this acoustic sensor system (S) for classification, with a) the output signals of the acoustic signal receiver (1) being fed to the classification module (8), b) the acoustic sensor system (S) as a recording system, implemented in Classification module (8) only for any acoustic signals of the stationary, different acoustic background of the sensor in preparation for a training phase for a new acoustic model or one to be updated, as well as a classification module (8) for known signal patterns during a classification phase and which is used for this necessary recognition acoustic models are generated from previously recorded signals which are the same for all sensors of the acoustic sensor system (S) and the respectively different acoustic background signals of the respective sensor, c) a user has the option of accessing the otherwise fully automatic process of acoustic model to influence generation by configuration of the acoustic sensor system (S), d) for model generation by the model update module (9) a corresponding recognition model is automatically generated and made available to the classification module (8) and e) each acoustic sensor system (S) uses the same recordings with regard to the acoustic signals to be recognized, but otherwise uses its own recordings with regard to the acoustic background, which differs for each sensor in the sensor system (S), and this model is then only used by this sensor in the sensor system (S) for classification. procedure after claim 1 , characterized in that an arranged model update module (9) for modeling is connected to the acoustic sensor system (S) and in that the management application (14) largely supports the user in monitoring and configuring the sensor of the acoustic sensor system (S ) supported, with a model update cycle within the model update module (9) based on data from the classification module (8) running automatically in a loop and the acoustic models created by this cycle by the respective addition of the stationary acoustic Background deviations from the hitherto known acoustic background of the respective sensor per sensor system (S) are formed and used by the sensor of the sensor system (S). procedure after claim 1 and 2 , characterized in that the acoustic sensor system (S) has a microcomputer, by means of which the sensor systems (S) are coupled to one another via the network (16) and the management application (14) and an evaluation application (13) and that the Classification module (8) with integrated acoustic background signal recording module and the model update module (9) from the management application (14) and the classification module (8) with integrated recording module if necessary from the model Update module (9) can be activated and deactivated. Device for performing automatic recognition and classification of acoustic signals according to claim 1 with at least one acoustic signal receiver (1) arranged in the monitoring area and a module for classifying the acoustic signal and outputting classification information, characterized in that the device has at least one acoustic sensor system (S), a raw acoustic database (10) of the acoustic sensor system ( 5), interfaces (S1 to S6) with an A/D converter (2), a hardware module motion detector (3), a hardware module GPS (4), a hardware module video camera (5), a hardware module wind sensor (6) and a hardware module rain sensor (7), an acoustic model database (11) in which acoustic models are stored and a classification module (8) with an integrated acoustic background signal recording module, which is connected to the acoustic model database (11) and to an evaluation application (13) and, if a signal is detected, the classification result to the evaluation application ation (13) so that the acoustic sensor system (5) serves both as a recording system for deviations in the acoustic background for a new training phase and as a classification module for known signal patterns during a classification phase and the necessary recognition models from the acoustic sensor system (S) previously recorded signals and different acoustic background data individually in each sensor system and these models are then only used by this acoustic sensor system (5) for classification. device after claim 4 characterized in that for the pre-processing of the acoustic signals an acoustic sensor system (5) having a microcomputer is provided, which is connected to the acoustic signal receiver (1), the output signals of which are fed to a classification module (8) which on the one hand the output signals classified and compared with existing acoustic models from the acoustic model database (11) and in the event of a match in the acoustic detection database (12 and/or 17) depending on the operating mode of the acoustic sensor system (S) for further processing and alerting and on the other hand carries out a permanent analysis of the stationary acoustic background and saves the acoustic deviations found in the raw acoustic database (10), whereupon after a configurable number of such deviations in the acoustic background the model update module (9) automatically and without required Action by the user the acoustic Mo Dell database (11) updated, after which the classification module (8) works with the currently prevailing acoustic conditions of the acoustic background and henceforth carries out its classifications with this updated acoustic model. device after claim 4 , characterized in that for the pre-processing of the acoustic signals a microcomputer (S-MC) having acoustic sensor system (5) is provided, which is connected to the acoustic signal receiver (1) whose output signals are fed to a classification module (8). which on the one hand classifies the output signals and compares them with existing acoustic models from the acoustic model database (11) and stores them in the acoustic recognition database (12 or 17) depending on the operating mode for further processing and alarming if they match, and on the other hand carries out a permanent analysis of the stationary acoustic background and stores the detected acoustic deviations in the acoustic detection database (12), whereupon after a configurable number of such deviations in the acoustic background the model update module (9) automatically and without any action required by the user the acoustic model database (11) ak tualized, following which the classification module (8) works with the currently prevailing acoustic conditions of the acoustic background and henceforth carries out its classifications with this updated acoustic model and the model update cycle within the model update module (9). based on data from the classification module (8) runs automatically and in a loop. device after claim 4 , characterized in that the model update module (9) for modeling and model updating with the acoustic sensor systems (S) is connected and which the classification module (8) both during an in-process acoustic model update regarding. Further acoustic Recordings of deviations in the acoustic background of the respective sensor, as well as after the acoustic model update has been activated and deactivated. device after claim 4 until 7 , characterized in that the acoustic sensor system (S) has an A/D converter (2) with an integrated amplifier connected to the acoustic signal receiver (1), whose amplified audio output signal is fed to an A/D converter (2). and that a microcomputer (S-MC) both with the A/D converter (2) via the interface (S1) and via at least one of the interfaces (S2, S3, S4, S5, S6) with at least one hardware module -Motion detector (3), a hardware module GPS (4), a hardware module video camera (5), a hardware module wind sensor (6) and/or a hardware module rain sensor (7) for Data exchange and for controlling the sensor system (S) via the computer network (16) is connected. device after claim 4 until 8th , characterized in that a central unit (Z) via a computer network (16) is connected to the sensor system (S) and the central unit (Z) contains a central microcomputer (Z-MC), a management application (14), a Evaluation application (13) and an acoustics detection database (17), wherein the central microcomputer (Z-MC) has a hardware module power supply (15), a hardware module wind sensor (6) and a hardware module Rain sensor (7) is connected.

Images